In Scott's latest post, he mentions OpenAI's charter section that states:
"Therefore, if a value-aligned, safety-conscious project comes close to building AGI before we do, we commit to stop competing with and start assisting this project. We will work out specifics in case-by-case agreements, but a typical triggering condition might be “a better-than-even chance of success in the next two years.”"
He then points out that there is a lot of ambiguity in decisionmaking. This market is intended to capture the chance of OpenAI's actions on that commitment differing from Scott's beliefs about how that commitment should operate.
YES if Scott makes such a statement, then OpenAI is assisting another company's efforts while not competing within 3 months. It also counts if Scott approves of OpenAI doing so after they've already started.
NO if Scott makes such a statement, and there is no public evidence of OpenAI assisting another AGI project and ceasing competition.
N/A if Scott never makes such a statement before market close (2050).
Note that OpenAI's assistance of that other project doesn't have to be at all caused by Scott's statement, just happen within 3 months of it.
Context: https://astralcodexten.substack.com/p/openais-planning-for-agi-and-beyond
Given that there's some ambiguity in the resolution here, I won't bet.
Mar 1, 10:38pm: If Scott Alexander publicly states that he believes OpenAI should activate the "long-term safety" part of their charter, will OpenAI do so within 3 months? → If Scott Alexander publicly states that he believes OpenAI should activate the "long-term safety" part of their charter, will OpenAI do so within 3 months?
1,000People are also trading
<deleted>
This apparently has been too confusing, so I’m just resolving N/A. The idea of the market was basically p(OpenAI does it | Scott thinks they should), but I can see how it didn’t come across. “Have done so” instead of “do so” didn’t fit in the character limit lol, but suspect that’s not the only problem.
Someone with a better phrasing can steal the idea or suggest it.
@DanStoyell Maybe have a market on p(OpenAI does it) and a market on P(Scott thinks they should do it) and see how they align.
@Odoacre Sorry if it's not clear via the fine print, but there is no required order. It accounts even if Scott approves of this action after OpenAI takes it.
@DanStoyell thanks for the clarification, that makes a big difference. I'm still happy with NO though
Scott Alexander is not an AI expert and does not have any special non-public information that gives him a special insight into progress towards AGI, as far as I know. Why, other than the possibility that a bunch of OpenAI engineers read his blog, should OpenAI particularly care what some blogger who happens to have a day job as a psychiatrist thinks?!
@RobinGreen
> other than the possibility that a bunch of OpenAI engineers read his blog
that does seem to be a significant possibility. Another possibility is a wider or broader pattern of influence by Scott on people engaged in debate, who influence OpenAI.
I think the most likely way this would turn out to be true, though, is that at some point in time, there are signs that the trigger condition will be met, and many people see this, including Scott, who writes about it, as well as OpenAI decision-makers, who independently decide to pull the trigger.
@B Agree. This market is intended to be more about the likelihood of agreement than about Scott causing OpenAI to act a certain way.